Active Machine Learning-driven Experiments on Malaria Cell Classification
Mingyong Ma · 2021 IEEE 3rd International Conference on Frontiers Technology of Information and Computer (ICFTIC) · 2021
Malaria is a life-threatening disease caused by Plasmodium parasites that infect the red blood cells. Computer-aided diagnostic(CADx) tools are implemented for identifying Malaria disease. However, it is not satisfying due to limited data. Therefore, in this literature, a pool-based active learning strategy is implemented to solve the problem. Logistic Regression and SVM are used to find out if active learning(AL) do makes the AL smarter and a Pre-trained resnet 50 is used to see if AL is a time-efficient way of training. It is observed that with 26% of the entire dataset, the accuracy reaches 93%. AL using uncertainty sampling has a better performance than random sampling. The time cost ratio determines the performance of three sampling methods under the same time cost.